coinductive predicates and
play

Coinductive Predicates and Final Sequences in a Fibration Ichiro - PowerPoint PPT Presentation

Coinductive Predicates and Final Sequences in a Fibration Ichiro Hasuo Bart Jacobs Kenta Cho Toshiki Kataoka Radboud Univ. Nijmegen (NL) University of Tokyo (JP) MFPS (Tulane) 2013/ 6/24 Coinduction Hasuo (Tokyo) O / / O


  1. Establish/Compute/Construct Coinductive Predicates ( c − 1 � ϕ � ) X ( c − 1 � ϕ � ) 2 X � � � X · · · � 0 � � 1 � � 2 � Hasuo (Tokyo)

  2. Establish/Compute/Construct Coinductive Predicates ( c − 1 � ϕ � ) X ( c − 1 � ϕ � ) 2 X � � � X · · · � | � n � , but � | = ν u. � u = n � 0 � � 1 � � 2 � Hasuo (Tokyo)

  3. Establish/Compute/Construct Coinductive Predicates ( c − 1 � ϕ � ) X ( c − 1 � ϕ � ) 2 X � � � X · · · � | � n � , but � | = ν u. � u = n � 0 � � 1 � � 2 � State space bound [Cousot & Cousot, ’79] | X | steps Hasuo (Tokyo)

  4. Establish/Compute/Construct Coinductive Predicates ( c − 1 � ϕ � ) X ( c − 1 � ϕ � ) 2 X � � � X · · · � | � n � , but � | = ν u. � u = n � 0 � � 1 � � 2 � State space bound [Cousot & Cousot, ’79] | X | steps “Behavioral bound” [Hennessy & Milner, ’85] ω steps if finitely branching ! Hasuo (Tokyo)

  5. Behavioral Bound for Computing Coind. Pred. P ( X ) Theorem. Let a Kripke frame be finitely branching. Then c � X ( c − 1 � ϕ � ) X ( c − 1 � ϕ � ) 2 X X � � � · · · stabilizes after ω steps. Hasuo (Tokyo)

  6. Behavioral Bound for Computing Coind. Pred. P ( X ) Theorem. Let a Kripke frame be finitely branching. Then c � X ( c − 1 � ϕ � ) X ( c − 1 � ϕ � ) 2 X X � � � · · · stabilizes after ω steps. Proof : Suffices to show is an invariant. � � n � n Hasuo (Tokyo)

  7. Behavioral Bound for Computing Coind. Pred. P ( X ) Theorem. Let a Kripke frame be finitely branching. Then c � X ( c − 1 � ϕ � ) X ( c − 1 � ϕ � ) 2 X X � � � · · · stabilizes after ω steps. x Proof : Suffices to show ⋯ is an invariant. � x 1 x 2 x k � n � n Hasuo (Tokyo)

  8. Behavioral Bound for Computing Coind. Pred. P ( X ) Theorem. Let a Kripke frame be finitely branching. Then c � X ( c − 1 � ϕ � ) X ( c − 1 � ϕ � ) 2 X X � � � · · · stabilizes after ω steps. x � � n � Proof : Suffices to show n ⋯ is an invariant. � x 1 x 2 x k � n � n Hasuo (Tokyo)

  9. Behavioral Bound for Computing Coind. Pred. P ( X ) Theorem. Let a Kripke frame be finitely branching. Then c � X ( c − 1 � ϕ � ) X ( c − 1 � ϕ � ) 2 X X � � � · · · stabilizes after ω steps. x � � n � Proof : Suffices to show n ⋯ is an invariant. � x 1 x 2 x k � n � n � 0 � Hasuo (Tokyo)

  10. Behavioral Bound for Computing Coind. Pred. P ( X ) Theorem. Let a Kripke frame be finitely branching. Then c � X ( c − 1 � ϕ � ) X ( c − 1 � ϕ � ) 2 X X � � � · · · stabilizes after ω steps. x � � n � Proof : Suffices to show n ⋯ is an invariant. � x 1 x 2 x k � n � n � 1 � � 0 � Hasuo (Tokyo)

  11. Behavioral Bound for Computing Coind. Pred. P ( X ) Theorem. Let a Kripke frame be finitely branching. Then c � X ( c − 1 � ϕ � ) X ( c − 1 � ϕ � ) 2 X X � � � · · · stabilizes after ω steps. x � � n � Proof : Suffices to show n ⋯ is an invariant. � x 1 x 2 x k � n � n � 1 � � 0 � � 2 � Hasuo (Tokyo)

  12. Behavioral Bound for Computing Coind. Pred. P ( X ) Theorem. Let a Kripke frame be finitely branching. Then c � X ( c − 1 � ϕ � ) X ( c − 1 � ϕ � ) 2 X X � � � · · · stabilizes after ω steps. x � � n � Proof : Suffices to show n ⋯ is an invariant. � x 1 x 2 x k � n � n � 1 � � 0 � � 3 � � 2 � Hasuo (Tokyo)

  13. Behavioral Bound for Computing Coind. Pred. P ( X ) Theorem. Let a Kripke frame be finitely branching. Then c � X ( c − 1 � ϕ � ) X ( c − 1 � ϕ � ) 2 X X � � � · · · stabilizes after ω steps. x � � n � Proof : Suffices to show n ⋯ is an invariant. � x 1 x 2 x k � n � n � 1 � � 0 � � 3 � � 2 � ⋮ ⋮ ⋮ Hasuo (Tokyo)

  14. Behavioral Bound for Computing Coind. Pred. P ( X ) Theorem. Let a Kripke frame be finitely branching. Then c � X ( c − 1 � ϕ � ) X ( c − 1 � ϕ � ) 2 X X � � � · · · stabilizes after ω steps. x � � n � Proof : Suffices to show n ⋯ is an invariant. � x 1 x 2 x k � n � n � 1 � � 0 � � 3 � = � n � � i � [1 , k ] s.t. x i | � 2 � ⋮ ⋮ for infinitely many n ⋮ Hasuo (Tokyo)

  15. Coinductive Predicates Conventionally (Summary) Hasuo (Tokyo)

  16. Coinductive Predicates Conventionally (Summary) (current st.) ⊨ P witnesses (next st.) ⊨ P Hasuo (Tokyo)

  17. Coinductive Predicates Conventionally (Summary) (current st.) ⊨ P witnesses (next st.) ⊨ P � � c − 1 ϕ � 2 X � 2 P X � 2 X � ν u. � u � = gfp P ( X ) c � X Hasuo (Tokyo)

  18. Coinductive Predicates Conventionally (Summary) (current st.) ⊨ P witnesses (next st.) ⊨ P ?? � � c − 1 ϕ � 2 X � 2 P X � 2 X � ν u. � u � = gfp P ( X ) c � X Hasuo (Tokyo)

  19. Coinductive Predicates Conventionally (Summary) (current st.) ⊨ P witnesses (next st.) ⊨ P ?? � � c − 1 ϕ � 2 X � 2 P X � 2 X � ν u. � u � = gfp P ( X ) c � X By Knaster-Tarski Hasuo (Tokyo)

  20. Coinductive Predicates Conventionally (Summary) (current st.) ⊨ P witnesses (next st.) ⊨ P ?? � � c − 1 ϕ � 2 X � 2 P X � 2 X � ν u. � u � = gfp P ( X ) c � X By Knaster-Tarski Inductive constr. ( c − 1 � ϕ � ) X ( c − 1 � ϕ � ) 2 X X � � � · · · Hasuo (Tokyo)

  21. Coinductive Predicates Conventionally (Summary) (current st.) ⊨ P witnesses (next st.) ⊨ P ?? � � c − 1 ϕ � 2 X � 2 P X � 2 X � ν u. � u � = gfp P ( X ) c � X By Knaster-Tarski Inductive constr. ( c − 1 � ϕ � ) X ( c − 1 � ϕ � ) 2 X X � � � · · · State space bound vs. “ behavioral bound ” Hasuo (Tokyo)

  22. Coinductive Predicates Conventionally (Summary) (current st.) ⊨ P witnesses (next st.) ⊨ P ?? � � c − 1 ϕ � 2 X � 2 P X � 2 X � ν u. � u � = gfp P ( X ) c � X By Knaster-Tarski Inductive constr. ( c − 1 � ϕ � ) X ( c − 1 � ϕ � ) 2 X X � � � · · · current work State space bound vs. “ behavioral bound ” Hasuo (Tokyo) current work

  23. Part II: Coinductive Predicates, Categorically

  24. Contributions Sufficient condition for categorical behavioral ω -bound based on F X P c � Coalgebra (transition system) ↓ p X C Fibration (underlying logic) ϕ P P � � p Predicate lifting (modality) p � � � � C C � � Y F � Locally presentable category (“size”) . X � m . i i ∃ l o C i . � � � � X ∃ i . . . Hasuo (Tokyo)

  25. Contributions Sufficient condition for categorical behavioral ω -bound based on F X P c � Coalgebra (transition system) ↓ p X C Fibration (underlying logic) ϕ P P � � p Predicate lifting (modality) p � � � � C C � � Y F � Locally presentable category (“size”) . X � m . i i ∃ l o C i . � � � � X ∃ i . . . Constr. of final coalg. by final sequence [Worrell, Adamek] Hasuo (Tokyo)

  26. Contributions Sufficient condition for categorical behavioral ω -bound based on F X P c � Coalgebra (transition system) ↓ p X C Fibration (underlying logic) ϕ P P � � p Predicate lifting (modality) p � � � � C C � � Y F � Locally presentable category (“size”) . X � m . i i ∃ l o C i . � � � � X ∃ i . . . Constr. of final coalg. by Coind. predicate as a final sequence [Worrell, Adamek] final coalgebra [Hermida, Jacobs] Hasuo (Tokyo)

  27. Categorical infrastructure: Contributions fibration and locally presentable cat. Sufficient condition for categorical behavioral ω -bound based on F X P c � Coalgebra (transition system) ↓ p X C Fibration (underlying logic) ϕ P P � � p Predicate lifting (modality) p � � � � C C � � Y F � Locally presentable category (“size”) . X � m . i i ∃ l o C i . � � � � X ∃ i . . . Constr. of final coalg. by Coind. predicate as a final sequence [Worrell, Adamek] final coalgebra [Hermida, Jacobs] Hasuo (Tokyo)

  28. Categorical infrastructure: Some math work Contributions fibration and locally presentable cat. Sufficient condition for categorical behavioral ω -bound based on F X P c � Coalgebra (transition system) ↓ p X C Fibration (underlying logic) ϕ P P � � p Predicate lifting (modality) p � � � � C C � � Y F � Locally presentable category (“size”) . X � m . i i ∃ l o C i . � � � � X ∃ i . . . Constr. of final coalg. by Coind. predicate as a final sequence [Worrell, Adamek] final coalgebra [Hermida, Jacobs] Hasuo (Tokyo)

  29. The Categorical Setup P ω ( X ) Kripke model c O X coinductive ν u. � u specification ϕ 3 : 2 X − → 2 P ω X { X � � X | U �� X � � U � = � } / 2 P X c − 1 / 2 X ϕ 3 2 X monotone ( c − 1 � ϕ 3 ) U ✓ invariant U Hasuo (Tokyo)

  30. The Categorical Setup P ω ( X ) F X Kripke model coalgebra c O c O X X coinductive ν u. � u specification ϕ 3 : 2 X − → 2 P ω X { X � � X | U �� X � � U � = � } / 2 P X c − 1 / 2 X ϕ 3 2 X monotone ( c − 1 � ϕ 3 ) U ✓ invariant U Hasuo (Tokyo)

  31. The Categorical Setup finitely branching P ω ( X ) F X Kripke model coalgebra c O c O X X coinductive ν u. � u specification ϕ 3 : 2 X − → 2 P ω X { X � � X | U �� X � � U � = � } / 2 P X c − 1 / 2 X ϕ 3 2 X monotone ( c − 1 � ϕ 3 ) U ✓ invariant U Hasuo (Tokyo)

  32. The Categorical Setup finitely finitary branching P ω ( X ) F X Kripke model coalgebra c O c O X X coinductive ν u. � u specification ϕ 3 : 2 X − → 2 P ω X { X � � X | U �� X � � U � = � } / 2 P X c − 1 / 2 X ϕ 3 2 X monotone ( c − 1 � ϕ 3 ) U ✓ invariant U Hasuo (Tokyo)

  33. The Categorical Setup finitely finitary branching P ω ( X ) F X Kripke model coalgebra c O c O X X coinductive coinductive ν u. � u νϕ specification specification ϕ 3 : 2 X − → 2 P ω X { X � � X | U �� X � � U � = � } / 2 P X c − 1 / 2 X ϕ 3 2 X monotone ( c − 1 � ϕ 3 ) U ✓ invariant U Hasuo (Tokyo)

  34. The Categorical Setup finitely finitary branching P ω ( X ) F X Kripke model coalgebra c O c O X X coinductive coinductive ν u. � u νϕ specification specification ϕ 3 : 2 X − → 2 P ω X ϕ : P X − → P F X { X � � X | predicate lifting U �� X � � U � = � } / 2 P X c − 1 / 2 X ϕ 3 2 X monotone ( c − 1 � ϕ 3 ) U ✓ invariant U Hasuo (Tokyo)

  35. The Categorical Setup finitely finitary branching P ω ( X ) F X Kripke model coalgebra c O c O X X coinductive coinductive ν u. � u νϕ specification specification ϕ 3 : 2 X − → 2 P ω X ϕ : P X − → P F X { X � � X | predicate lifting U �� X � � U � = � } / 2 P X c − 1 / 2 X c ∗ ϕ X / P F X ϕ 3 2 X P X / P X monotone endofunctor ( c − 1 � ϕ 3 ) U ✓ invariant U Hasuo (Tokyo)

  36. O The Categorical Setup finitely finitary branching P ω ( X ) F X Kripke model coalgebra c O c O X X coinductive coinductive ν u. � u νϕ specification specification ϕ 3 : 2 X − → 2 P ω X ϕ : P X − → P F X { X � � X | predicate lifting U �� X � � U � = � } / 2 P X c − 1 / 2 X c ∗ ϕ X / P F X ϕ 3 2 X P X / P X monotone endofunctor ( c − 1 � ϕ 3 ) U ( c ∗ � ϕ ) P coalgebra ✓ invariant (in a fibr.) U P Hasuo (Tokyo)

  37. X O coinductive coinductive ν u. � u νϕ specification specification ϕ 3 : 2 X − → 2 P ω X ϕ : P X − → P F X { X � � X | predicate lifting U �� X � � U � = � } / 2 P X c − 1 / 2 X c ∗ ϕ X / P F X ϕ 3 2 X P X / P X monotone endofunctor ( c − 1 � ϕ 3 ) U ( c ∗ � ϕ ) P coalgebra ✓ invariant (in a fibr.) U P ( c − 1 � ϕ 3 ) J ν u. ϕ 3 u K c coind. pred. J ν u. ϕ 3 u K c X ◆ ( c − 1 � ϕ 3 ) X ◆ · · · inductive constr. Hasuo (Tokyo)

  38. X O coinductive coinductive ν u. � u νϕ specification specification ϕ 3 : 2 X − → 2 P ω X ϕ : P X − → P F X { X � � X | predicate lifting U �� X � � U � = � } / 2 P X c − 1 / 2 X c ∗ ϕ X / P F X ϕ 3 2 X P X / P X monotone endofunctor ( c − 1 � ϕ 3 ) U ( c ∗ � ϕ ) P coalgebra ✓ invariant (in a fibr.) U P ( c − 1 � ϕ 3 ) J ν u. ϕ 3 u K c ( c ∗ � ϕ ) J νϕ K c final coalg. ⇠ = O coind. pred. (in a fibr.) J ν u. ϕ 3 u K c J νϕ K c X ◆ ( c − 1 � ϕ 3 ) X ◆ · · · inductive constr. Hasuo (Tokyo)

  39. X O coinductive coinductive ν u. � u νϕ specification specification ϕ 3 : 2 X − → 2 P ω X ϕ : P X − → P F X { X � � X | predicate lifting U �� X � � U � = � } / 2 P X c − 1 / 2 X c ∗ ϕ X / P F X ϕ 3 2 X P X / P X monotone endofunctor ( c − 1 � ϕ 3 ) U ( c ∗ � ϕ ) P coalgebra ✓ invariant (in a fibr.) U P ( c − 1 � ϕ 3 ) J ν u. ϕ 3 u K c ( c ∗ � ϕ ) J νϕ K c final coalg. ⇠ = O coind. pred. (in a fibr.) J ν u. ϕ 3 u K c J νϕ K c X ◆ ( c − 1 � ϕ 3 ) X ◆ · · · > X ( c ∗ � ϕ X ) > X · · · inductive constr. final sequence in a fibr. Hasuo (Tokyo)

  40. What Categorical Generalization Buys Us Hasuo (Tokyo)

  41. What Categorical Generalization Buys Us Final coalgebra in C : (strongly) LFP ( Posets , Graphs , Vec , ...) [Adamek ’03] Coinductive pred. for different F : Sets → Sets Coalg. μ -calculus ; coalg. automata [Cirstea, Kupke & Pattinson, CSL ’09] [Cirstea & Sadrzadeh, CMCS’08] [Venema, I&C’06] Hasuo (Tokyo)

  42. What Categorical Generalization Buys Us Final coalgebra in C : (strongly) LFP ( Posets , Graphs , Vec , ...) [Adamek ’03] Coinductive pred. for different F : Sets → Sets Coalg. μ -calculus ; coalg. automata [Cirstea, Kupke & Pattinson, CSL ’09] [Cirstea & Sadrzadeh, CMCS’08] [Venema, I&C’06] P Various “underlying logics” as ↓ p C Sub( C ) Rel ↓ Sub(Sets F ) ( C : a topos) ↓ C ↓ Sets Constructive logics Sets F Relations (“binary pred. ”) For name-passing Hasuo (Tokyo)

  43. What Categorical Generalization Buys Us Final coalgebra in C : (strongly) LFP ( Posets , Graphs , Vec , ...) [Adamek ’03] Coinductive pred. for different F : Sets → Sets Coalg. μ -calculus ; coalg. automata [Cirstea, Kupke & Pattinson, CSL ’09] [Cirstea & Sadrzadeh, CMCS’08] [Venema, I&C’06] P Coind. relations Various “underlying logics” as ↓ p e.g. bisimilarity C Sub( C ) Rel ↓ Sub(Sets F ) ( C : a topos) ↓ C ↓ Sets Constructive logics Sets F Relations (“binary pred. ”) For name-passing Hasuo (Tokyo)

  44. Coindution in a Fibration P Rel Pred conventional relational fibrational ↓ p ↓ ↓ C Sets Sets invariant bisimulation coalgebra coind. pred. bisimilarity final coalg. partition inductive constr. final sequence refinement Hasuo (Tokyo)

  45. Part III: Technical Ingredients Final Sequence, Fibration, Predicate Lifting, Locally Finitely Presentable Category, ...

  46. [Worrell, TCS’05] in Sets o o o o [Adamek, TCS’03] in strongly LFP C Final Sequence F i 1 1 F 1 ! · · · · · · Hasuo (Tokyo)

  47. [Worrell, TCS’05] in Sets o o o o [Adamek, TCS’03] in strongly LFP C Final Sequence { i-step behaviors} F i 1 1 F 1 ! · · · · · · Hasuo (Tokyo)

  48. [Worrell, TCS’05] in Sets o s t { o o o [Adamek, TCS’03] in strongly LFP C Final Sequence lim { i-step behaviors} F ω 1 π i F i 1 1 F 1 ! · · · · · · Hasuo (Tokyo)

  49. [Worrell, TCS’05] in Sets k d s s o y _ j o o � o [Adamek, TCS’03] in strongly LFP C Final Sequence lim { i-step behaviors} F ω 1 5 π i 1 F i 1 F 1 ! b · · · · · · F π i − 1 F ( F ω 1) Hasuo (Tokyo)

  50. [Worrell, TCS’05] in Sets o k j o o o d _ y s s � [Adamek, TCS’03] in strongly LFP C Final Sequence lim { i-step behaviors} F ω 1 5 π i 1 F i 1 F 1 ! b · · · · · · F π i − 1 F ( F ω 1) : a final coalgebra? F ω 1 Yes, when F is limit preserving ( b is iso) Hasuo (Tokyo)

  51. [Worrell, TCS’05] in Sets _ s o o k j y o s � d o [Adamek, TCS’03] in strongly LFP C Final Sequence lim { i-step behaviors} F ω 1 5 π i 1 F i 1 F 1 ! b · · · · · · F π i − 1 F ( F ω 1) : a final coalgebra? F ω 1 Yes, when F is limit preserving ( b is iso) Almost, when F is finitary ( b is monic) Quotient modulo beh. eq. Continue till ω + ω [Worrell] Hasuo (Tokyo)

  52. P Fibration ↓ p C f / Y in C X “Organize indexed entities,” categorically In particular: categorical model of predicate logics Hasuo (Tokyo)

  53. P Fibration ↓ p C f / Y in C X indices “Organize indexed entities,” categorically In particular: categorical model of predicate logics Hasuo (Tokyo)

  54. P Fibration ↓ p P X C P ’ P ’’ P f / Y in C X indices “Organize indexed entities,” categorically In particular: categorical model of predicate logics Hasuo (Tokyo)

  55. P Fibration ↓ p P X P Y C Q ’ P ’ P ’’ Q P f / Y in C X indices “Organize indexed entities,” categorically In particular: categorical model of predicate logics Hasuo (Tokyo)

  56. P Fibration ↓ p P X P Y C Q ’ P ’ P ’’ indexed entities Q P f / Y in C X indices “Organize indexed entities,” categorically In particular: categorical model of predicate logics Hasuo (Tokyo)

  57. o P Fibration ↓ p P X P Y C Q ’ P ’ f ∗ P ’’ indexed entities Q P f / Y in C X indices “Organize indexed entities,” categorically In particular: categorical model of predicate logics Hasuo (Tokyo)

  58. o P Fibration ↓ p “substitution” P X P Y C Q ’ P ’ f ∗ P ’’ indexed entities Q P f / Y in C X indices “Organize indexed entities,” categorically In particular: categorical model of predicate logics Hasuo (Tokyo)

  59. o P Fibration ↓ p “substitution” P X P Y C Q ’ P ’ f ∗ P ’’ indexed entities Q P f / Y in C X indices “Organize indexed entities,” categorically In particular: categorical model of predicate logics Hasuo (Tokyo)

  60. o o P Fibration ↓ p “substitution” P X P Y C Q ’ P ’ f ∗ P ’’ indexed entities Q P f / Y in C X indices “Organize indexed entities,” categorically : predicates over X ( P X, ⊆ ) In particular: Substitution f − 1 ( V ⊆ Y ) categorical model of f − 1 P X P Y � � = V f ( ) predicate logics f / Y X Hasuo (Tokyo)

  61. o Fibration: from Pointwise Indexing to Display Indexing P X P Y Q ’ f ∗ P ’ P ’’ Q P f / Y X in C Hasuo (Tokyo)

Recommend


More recommend